AMAPHUZU AYINHLOKO
- Ama-Tech giants agijima ukuze aphushele amathuluzi azo akhiqizayo e-AI aye ezingeni elilandelayo.
- Abanye ochwepheshe bathi ubuchwepheshe bungakhulula abantu emisebenzini ephansi futhi bungenise inkathi entsha yokudala.
- Kodwa kungase kubeke engcupheni imisebenzi futhi kubangele izinkinga zenhlalo ezingenakuxazululeka.
Ukunyuka kwe – manje ebonwa njengento engenakugwenywa eSilicon Valley – izoletha ushintsho “oluyimiyalo yobukhulu” enkulu kunanoma yini umhlaba ongakayiboni, kusho izingqapheli. Kodwa ingabe sesilungile?
I-AGI – ichazwa njengobuhlakani bokwenziwa obunamakhono okuqonda omuntu, ngokuphambene nobuhlakani bokwenziwa obuncane kakhulu, njengokuthi – ingakhulula abantu emisebenzini ephansi futhi ingenise inkathi entsha yokuqamba.
Kodwa ushintsho oluyingqophamlando olunjalo lwepharadigm lungaphinde lufake engcupheni imisebenzi futhi luphakamise izinkinga zenhlalo ezingenakuxazululeka, kuxwayisa ochwepheshe.
Intuthuko yangaphambilini yezobuchwepheshe kusukela kugesi kuya ku-inthanethi ibangele ushintsho olunamandla emphakathini, kusho uSiqi Chen, isikhulu esiphezulu se-San Francisco startup Runway.
“Kodwa esikubhekile manje wubuhlakani uqobo lwayo… Kungokokuqala ngqa ukuthi sikwazi ukuzakhela ezobunhloli futhi sandise inani lazo endaweni yonke,” etshela abezindaba ze-AFP.
I-OpenAI ithi i-ChatGPT-4 yayo entsha “ibonisa ukusebenza kwezinga lomuntu” kwamanye amabhentshimakhi. Umthombo: I-AAP / URichard Drew/AP
Ushintsho, ngenxa yalokho, luyoba “ama-oda obukhulu obukhulu kunawo wonke ushintsho lwezobuchwepheshe esake saba nalo emlandweni.”
Futhi uguquko olunjalo olujabulisayo, nolwesabekayo “luyinkemba esika nhlangothi zombili,” kusho uMnu Chen, ecabanga ukusebenzisa i-AGI ukubhekana nokushintsha kwesimo sezulu, ngokwesibonelo, kodwa futhi exwayisa ngokuthi iyithuluzi esifuna ukuba “liqine ngangokunokwenzeka. “
Kwaba ukukhululwa kwe ngasekupheleni konyaka odlule okwaletha umbono wephupho elide le-AGI one giant jump eduzane neqiniso.
I-OpenAI, inkampani ngemuva kwesoftware ekhiqizayo izinkondlo, nekhodi yekhompuyutha ngomyalo, kuleli sonto kukhishwe inguqulo enamandla nakakhulu yobuchwepheshe obusebenzayo – GPT-4.
Ithe ubuchwepheshe ngeke bukwazi ukucubungula umbhalo kuphela kodwa nezithombe, futhi bukhiqize okuqukethwe okuyinkimbinkimbi njengezikhalazo zomthetho noma imidlalo yevidiyo.
Ngakho-ke “ikhombisa ukusebenza kwezinga labantu” kwamanye amabhentshimakhi, kusho inkampani.
Ingabe i-AI ingasho ukuphela ‘kokudlova’?
Impumelelo ye-OpenAI, esekelwa yiMicrosoft, idale umjaho wezikhali eSilicon Valley njengoba iziqhwaga zobuchwepheshe zifuna ukuphusha amathuluzi azo akhiqizayo e-AI ziye ezingeni elilandelayo – yize zisaxwayile ama-chatbots aphuma ujantshi.
Kakade, abasizi bedijithali abafakwe nge-AI abavela ku-Microsoft kanye ne-Google bangafingqa imihlangano, babhale ama-imeyili, benze amawebhusayithi, imikhankaso yezikhangiso zobuciko, nokunye – okusinikeza amazwibela alokho i-AGI ezokwazi ukukwenza esikhathini esizayo.
“Sichitha isikhathi esiningi sisebenzisa isidina,” kusho uJared Spataro, iphini likamongameli webhizinisi leMicrosoft.
Ngobuhlakani bokwenziwa uMnu Spataro ufuna “ukuthola kabusha umphefumulo womsebenzi,” esho ngesikhathi esethula iMicrosoft ngoLwesine.
I-Artificial intelligence nayo inganciphisa izindleko, abanye basikisela.
Umakhi wezakhiwo waseBrithani uJoe Perkins ubhale kuTwitter ukuthi wasebenzisa i-GPT-4 ukwenza iphrojekthi yokubhala amakhodi, umthuthukisi “omuhle kakhulu” amtshele ukuthi izobiza opondo abangu-5,000 ($6,000) futhi ithathe amasonto amabili.
“I-GPT-4 ilethe okufanayo emahoreni angu-3, ngo-$0.11,” ubhale kuTwitter. “Iyamangaza ngempela.”
Kodwa lokho kuphakamisa umbuzo wosongo emisebenzini yabantu, usomabhizinisi u-Chen evuma ukuthi ubuchwepheshe ngolunye usuku bungakha isiqalo esifana nesakhe – noma inguqulo engcono nakakhulu.
“Ngizoziphilisa kanjani ngingabe ngingenakhaya?” ebuza, wengeza ngokuthi usethembele ezisombululweni ezizovela.
Imibuzo ekhona
Ubuhlakani bokwenziwa obutholakala yonke indawo buphinde bubeke uphawu lombuzo phezu kobuqiniso bokudala njengoba izingoma, izithombe, ubuciko, nokuningi kukhishwa isofthiwe esikhundleni sabantu.
Ingabe abantu bazoyigwema imfundo, esikhundleni salokho bathembele kuma-software ukuba abacabangele?
Futhi, ubani okufanele amethembe ukwenza i-AI ingachemi, inembe, futhi ivumelane nezimo emazweni namasiko ahlukene?
I-AGI “mhlawumbe iza kithi ngokushesha kunalokho esingakucubungula,” kusho uSharon Zhou, umsunguli wenkampani ekhiqiza i-AI.
Ubuchwepheshe buphakamisa umbuzo okhona esintwini, utshele abe-AFP.
“Uma kuzoba khona into enamandla kunathi futhi ehlakaniphe kakhulu kunathi, kusho ukuthini lokho ngathi?” Kubuza uNksz Zhou.
“Kanti siyayibophela yini? Noma iyasibophela?”
Impumelelo ye-OpenAI, esekelwa yiMicrosoft, idale umjaho wezikhali eSilicon Valley njengoba iziqhwaga zobuchwepheshe zifuna ukuphusha amathuluzi azo akhiqizayo e-AI ziye ezingeni elilandelayo – yize zisaxwayile ama-chatbots aphuma ujantshi.
I-OpenAI ithe ihlela ukwakha i-AGI kancane kancane ngenhloso yokuzuzisa sonke isintu, kodwa ivumile ukuthi isoftware inamaphutha okuphepha.
Ukuphepha “kuyinqubo,” kusho usosayensi omkhulu we-OpenAI u-Ilya Sutskever engxoxweni abe nayo ne-MIT Technology Review, wengeza ngokuthi “kungafiseleka kakhulu” ukuthi izinkampani “ziqhamuke nohlobo oluthile lwenqubo oluvumela ukukhishwa kancane kwamamodeli nge lawa makhono angakaze abonwe nhlobo.”
Kodwa okwamanje, kusho uNksz Zhou, ukwehlisa ijubane akuyona ingxenye ye-ethos.
“Amandla agxile kulabo abakwazi ukwakha lezi zinto. Futhi bathatha izinqumo mayelana nalokhu, futhi bathambekele ekuhambeni ngokushesha,” esho.
I-oda lamazwe ngamazwe ngokwalo lingase libe sengozini, uyaphakamisa.
“Ingcindezi phakathi kwe-US neChina ibe nkulu,” kusho uNksz Zhou, wengeza ngokuthi umjaho wezobunhloli wokwenziwa ubiza inkathi yeMpi Yomshoshaphansi.
“Impela bukhona ubungozi nge-AGI ukuthi uma izwe elilodwa likuthola ngokushesha lokho, bazobusa?” uyabuza.
“Futhi ngakho-ke ngicabanga ukuthi ukwesaba, ungayeki ngoba asikwazi ukulahlekelwa.”
Riaan 
